Re: Dredging up bugs...

Frank Kotler wrote:
> "H. Peter Anvin" wrote:
>
> > Urk.  I just found not just one, but two bugs, related to the same file.
> >
> > The bugs are:
> >
> > #1. The optimizer doesn't converge; it will run an arbitrary number of
> > passes (if you let it, it'll run 1000 passes.)
>
>
> I can get it to assemble using "-O3", by defining "HEXDATE" and
> "DATE_STR" on the command line - not from the makefile. I suspect a
> problem with the way Nasm is picking up (or not) those defines from the
> Makefile(?). Nasm barfs up that "phase error" message as a "bonus" if
> *anything* goes wrong (a "mini-bug"). I don't think the optimizer
> actually fails to converge.

That's different from what I'm seeing.  I have run this from the command 
line as well as from the Makefile and am seeing the same behaviour.

What *does* make it work, however, is to change FILENAME_MAX and 
FILENAME_MAX_LG2 from equ's to %defines.  If I assemble with -O0, it 
complains that these are used before they are defined, which is incorrect.

> > #2. The sequencing of passes when this happens is such that a list file
> > is never produced.
>
> H.P.Recktenwald had a "patch" to emit a list file no matter what... As I
> recall, it involved not setting "terminate_after_phase"... emitted a bad
> output file, too. Just removing "remove(listname)" at line 295 of nasm.c
> seems *not* to do it, as I recall. Was a long time ago I looked at it.

In my case it emits an output file, but I have no way of knowing if it 
is any good.

        -hpa


-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now. 
http://productguide.itmanagersjournal.com/